Artificial Intelligence Software Vulnerabilities

Timeline

The analysis of the timeline helps to identify the required approach and handling of single vulnerabilities and vulnerability collections. This overview makes it possible to see less important slices and more severe hotspots at a glance. Initiating immediate vulnerability response and prioritizing of issues is possible.

Vendor

Identifying all affected vendors is a good starting point for an overview. This makes it possible to determine an homogeneous landscape or the most important hotspots in heterogeneous landscapes.

Product

Google TensorFlow227
Google Tensorflow55
TensorFlow34
Petwant PF-1038
Petwant Petalk AI8

Grouping vulnerabilities by products helps to get an overview. This makes it possible to determine an homogeneous landscape or the most important hotspots in heterogeneous landscapes.

Remediation

Official Fix322
Temporary Fix0
Workaround2
Unavailable2
Not Defined17

Vendors and researchers are eager to find countermeasures to mitigate security vulnerabilities. These can be distinguished between multiple forms and levels of remediation which influence risks differently.

Exploitability

High0
Functional0
Proof-of-Concept2
Unproven0
Not Defined341

Researcher and attacker which are looking for security vulnerabilities try to exploit them for academic purposes or personal gain. The level and quality of exploitability can be distinguished to determine simplicity and strength of attacks.

Access Vector

Not Defined0
Physical3
Local6
Adjacent3
Network331

The approach a vulnerability it becomes important to use the expected access vector. This is typically via the network, local, or physically even.

Authentication

Not Defined0
High0
Low104
None239

To exploit a vulnerability a certail level of authentication might be required. Vulnerabilities without such a requirement are much more popular.

User Interaction

Not Defined0
Required201
None142

Some attack scenarios require some user interaction by a victim. This is typical for phishing, social engineering and cross site scripting attacks.

C3BM Index

Our unique C3BM Index (CVSSv3 Base Meta Index) cumulates the CVSSv3 Meta Base Scores of all entries over time. Comparing this index to the amount of disclosed vulnerabilities helps to pinpoint the most important events.

CVSSv3 Base

≤10
≤21
≤31
≤47
≤5160
≤654
≤787
≤818
≤98
≤107

The Common Vulnerability Scoring System (CVSS) is an industry standard to define the characteristics and impacts of security vulnerabilities. The base score represents the intrinsic aspects that are constant over time and across user environments. Our unique meta score merges all available scores from different sources to aggregate to the most reliable result.

CVSSv3 Temp

≤10
≤21
≤31
≤48
≤5165
≤699
≤745
≤814
≤93
≤107

The Common Vulnerability Scoring System (CVSS) uses temp scores to reflect the characteristics of a vulnerability that may change over time but not across user environments. This includes reporting confidence, exploitability and remediation levels. We do also provide our unique meta score for temp scores, even though other sources rarely publish them.

VulDB

≤10
≤21
≤31
≤437
≤5178
≤612
≤788
≤812
≤97
≤107

The moderation team is always defining the base vector and base score for an entry. These and all other available scores are used to generate the meta score.

NVD

≤10
≤20
≤30
≤42
≤54
≤615
≤78
≤818
≤911
≤108

The National Vulnerability Database (NVD) is also defining CVSS vectors and scores. These are usually not complete and might differ from VulDB scores.

CNA

≤10
≤20
≤30
≤40
≤54
≤642
≤732
≤826
≤912
≤101

A CVE Numbering Authority (CNA) is responsible for assigning new CVE entries. They might also include a CVSS score. These are usually not complete and might differ from VulDB scores.

Vendor

≤10
≤20
≤30
≤40
≤50
≤60
≤70
≤80
≤90
≤100

Some vendors are willing to publish their own CVSS vectors and scores for vulnerabilities in their products. The coverage varies from vendor to vendor.

Research

≤10
≤20
≤30
≤40
≤50
≤60
≤70
≤80
≤90
≤100

There are sometimes also security researcher which provide their own CVSS vectors and scores for vulnerabilities they have found and published.

Exploit 0-day

<1k16
<2k11
<5k73
<10k149
<25k94
<50k0
<100k0
≥100k0

The moderation team is working with the threat intelligence team to determine prices for exploits. Our unique algorithm is used to identify the 0-day prices for an exploit, before it got distributed or became public. Calculated prices are aligned to prices disclosed by vulnerability broker and compared to prices we see on exploit markets.

Exploit Today

<1k201
<2k58
<5k82
<10k1
<25k1
<50k0
<100k0
≥100k0

The 0-day prices do not consider time-relevant factors. The today price does reflect price impacts like disclosure of vulnerability details, alternative exploits, availability of countermeasures. These dynamic aspects might decrease the exploit prices over time. Under certain circumstances this happens very fast.

Exploit Market Volume

Our unique calculation of exploit prices makes it possible to forecast the expected exploit market volume. The calculated prices for all possible 0-day expoits are cumulated for this task. Comparing the volume to the amount of disclosed vulnerabilities helps to pinpoint the most important events.

🔴 CTI Activities

Our unique Cyber Threat Intelligence aims to determine the ongoing research of actors to anticipiate their acitivities. Observing exploit markets on the Darknet, discussions of vulnerabilities on mailinglists, and exchanges on social media makes it possible to identify planned attacks. Monitored actors and activities are classified whether they are offensive or defensive. They are also weighted as some actors are well-known for certain products and technologies. And some of their disclosures might contain more or less details about technical aspects and personal context. The world map highlights active actors in real-time.

Affected Products (23): Amazon Echo, Amazon Echo Dot, Amazon Echo Show, Amazon Echo Spot, Artificial Intelligence Theme, Chain Sea AI Chatbot Backend, Chain Sea AI Chatbot System, Google TensorFlow, Google Tensorflow, Mycroft AI, Petalk Petalk AI, Petalk PF-103, Petwant Petalk AI, Petwant PF-103, Sofy.AI Plugin, SUSI.AI, TensorFlow, toucbase.ai, touchbase.ai, Webexceluk P.A.I.D, Xiaomi AI Speaker, Xiaomi AI speaker MDZ-25-DT, Yeelight Smart AI Speaker

PublishedBaseTempVulnerabilityProductExpRemCTICVE
05/22/20226.05.9Google TensorFlow tf.raw_ops.EditDistance out-of-bounds writeTensorFlowNot DefinedOfficial Fix0.03CVE-2022-29208
05/22/20224.34.2Google TensorFlow denial of serviceTensorFlowNot DefinedOfficial Fix0.05CVE-2022-29207
05/22/20224.34.2Google TensorFlow tf.raw_ops.SparseTensorDenseAdd null pointer dereferenceTensorFlowNot DefinedOfficial Fix0.02CVE-2022-29206
05/22/20224.34.2Google TensorFlow tf.compat.v1 uninitialized resourceTensorFlowNot DefinedOfficial Fix0.03CVE-2022-29205
05/22/20224.34.2Google TensorFlow tf.raw_ops.SpaceToBatchND integer overflowTensorFlowNot DefinedOfficial Fix0.04CVE-2022-29203
05/22/20226.46.3Google TensorFlow saved_model_cli code injectionTensorFlowNot DefinedOfficial Fix0.07CVE-2022-29216
05/22/20224.34.2Google TensorFlow tf.compat.v1.signal.rfft3d denial of serviceTensorFlowNot DefinedOfficial Fix0.00CVE-2022-29213
05/22/20224.34.2Google TensorFlow TFLite Model assertionTensorFlowNot DefinedOfficial Fix0.09CVE-2022-29212
05/22/20224.34.2Google TensorFlow type confusionTensorFlowNot DefinedOfficial Fix0.05CVE-2022-29209
05/22/20224.34.2Google TensorFlow tf.raw_ops.UnsortedSegmentJoin integer underflowTensorFlowNot DefinedOfficial Fix0.03CVE-2022-29204
05/22/20224.34.2Google TensorFlow tf.ragged.constant resource consumptionTensorFlowNot DefinedOfficial Fix0.02CVE-2022-29202
05/22/20224.34.2Google TensorFlow tf.raw_ops.QuantizedConv2D null pointer dereferenceTensorFlowNot DefinedOfficial Fix0.02CVE-2022-29201
05/22/20224.34.2Google TensorFlow tf.raw_ops.LSTMBlockCell denial of serviceTensorFlowNot DefinedOfficial Fix0.03CVE-2022-29200
05/22/20224.34.2Google TensorFlow tf.raw_ops.LoadAndRemapMatrix denial of serviceTensorFlowNot DefinedOfficial Fix0.03CVE-2022-29199
05/22/20224.34.2Google TensorFlow tf.raw_ops.SparseTensorToCSRSparseMatrix denial of serviceTensorFlowNot DefinedOfficial Fix0.04CVE-2022-29198
05/22/20224.34.2Google TensorFlow tf.raw_ops.UnsortedSegmentJoin denial of serviceTensorFlowNot DefinedOfficial Fix0.02CVE-2022-29197
05/22/20224.34.2Google TensorFlow tf.raw_ops.Conv3DBackpropFilterV2 denial of serviceTensorFlowNot DefinedOfficial Fix0.02CVE-2022-29196
05/22/20224.34.2Google TensorFlow tf.raw_ops.StagePeek denial of serviceTensorFlowNot DefinedOfficial Fix0.02CVE-2022-29195
05/22/20224.34.2Google TensorFlow tf.raw_ops.DeleteSessionTensor denial of serviceTensorFlowNot DefinedOfficial Fix0.04CVE-2022-29194
05/22/20224.34.2Google TensorFlow tf.raw_ops.TensorSummaryV2 denial of serviceTensorFlowNot DefinedOfficial Fix0.02CVE-2022-29193
05/22/20224.34.2Google TensorFlow tf.raw_ops.QuantizeAndDequantizeV4Grad denial of serviceTensorFlowNot DefinedOfficial Fix0.07CVE-2022-29192
05/22/20224.34.2Google TensorFlow tf.raw_ops.GetSessionTensor denial of serviceTensorFlowNot DefinedOfficial Fix0.03CVE-2022-29191
05/22/20224.34.2Google TensorFlow tf.histogram_fixed_width denial of serviceTensorFlowNot DefinedOfficial Fix0.05CVE-2022-29211
05/22/20224.34.2Google TensorFlow TensorKey heap-based overflowTensorFlowNot DefinedOfficial Fix0.00CVE-2022-29210
02/24/20225.55.3Amazon Echo Dot Skill/Bluetooth AvA neutralizationEcho DotNot DefinedNot Defined0.05CVE-2022-25809

318 more entries are not shown

Do you want to use VulDB in your project?

Use the official API to access entries easily!